[XPM] BFGMiner Branch.. on github

[paulthetafy]

This code isn't working "as-is". We are still trying to figure out how to get it working.

GBT, on the other hand, works fine without primecoind modification.

If you're implying that you have it working, please can you confirm this and share how you have managed it?

[kendog77]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

[swordfish6975]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

--no-getwork      or is that to obvious?

[paulthetafy]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

--no-getwork      or is that to obvious?

Tried that early on - causes a segmentation fault for me on startup

[swordfish6975]

ah true

[kendog77]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

--no-getwork      or is that to obvious?

I tried that as well, and while bfgminer does not fail for me, it will not connect to primecoind with the --no-getwork option.

[AgentME]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

Is it using getwork successfully to talk to primecoind?

[paulthetafy]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

Is it using getwork successfully to talk to primecoind?

well it appears to getwork but fails submissions, so no.  I also think Luke-jr confirmed that getwork wasn't working

[liteuser]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

--no-getwork      or is that to obvious?

I tried that as well, and while bfgminer does not fail for me, it will not connect to primecoind with the --no-getwork option.

I think there is a problem with blkmk_address_to_script function in libblkmaker/base58.c
I added case 23 and 83 (primecoin PUBKEY_ADDRESS and SCRIPT_ADDRESS), but after 4 hours I have got no blocks ...

[paulthetafy]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

--no-getwork      or is that to obvious?

I tried that as well, and while bfgminer does not fail for me, it will not connect to primecoind with the --no-getwork option.

I think there is a problem with blkmk_address_to_script function in libblkmaker/base58.c
I added case 23 and 83 (primecoin PUBKEY_ADDRESS and SCRIPT_ADDRESS), but after 4 hours I have got no blocks ...

I guess one of the issues we have is that it takes so damn long to find a block that testing changes is painfully slow.  I have a 32-core ec2 instance running but even that is taking several hours to find a block

[AgentME]

BFGMiner prime is definitely using getwork when talking to primecoind.

I'm still poking around the source code to see if there is anyway to force it to use GBT.

--no-getwork      or is that to obvious?

I tried that as well, and while bfgminer does not fail for me, it will not connect to primecoind with the --no-getwork option.

I think there is a problem with blkmk_address_to_script function in libblkmaker/base58.c
I added case 23 and 83 (primecoin PUBKEY_ADDRESS and SCRIPT_ADDRESS), but after 4 hours I have got no blocks ...

I guess one of the issues we have is that it takes so damn long to find a block that testing changes is painfully slow.  I have a 32-core ec2 instance running but even that is taking several hours to find a block

I tried testing code on primecoin testnet, but I couldn't connect to any testnet nodes.

[xyzzy099]

Client node has no API support yet. Feel free to propose an API format for e.g. getwork.

Based on the above quote from Sunny King, it seems that the client simply has no API support for miners at all at the moment.

I think that needs to be done first before any miner program, BFGMiner or otherwise, can mine Primecoins...  Or am I missing something here?

[dreamwatcher]

Client node has no API support yet. Feel free to propose an API format for e.g. getwork.

Based on the above quote from Sunny King, it seems that the client simply has no API support for miners at all at the moment.

I think that needs to be done first before any miner program, BFGMiner or otherwise, can mine Primecoins...  Or am I missing something here?

Not quite sure.

I know getwork gives no valuable information:

Code:

{
"midstate" : "7601d3f08f8dd8ac7a67141fa7f24493c6839606ecaf82c6a62a00ad3636923a",
"data" : "000000029ff9bbfa7b85551842ecb1e786bea0b096f92916652235784277e741ee56291e05e3040e4aa68f7c4a2d3f2a91252ad22b25c4799dc88e67f29884a3aee5927751e026a307e7ea1e00000000000000800000000000000000000000000000000000000000000000000000000000000000000000000000000080020000",
"hash1" : "00000000000000000000000000000000000000000000000000000000000000000000008000000000000000000000000000000000000000000000000000010000",
"target" : "000000001eea6700000000000000000000000000000000000000000000000000"
}

Blocktemplate also appears to be missing some key information:

Code:

{
"version" : 2,
"previousblockhash" : "f104ee5b36efa6e6d1127306a14c7fe76253a3baaacba3504100533ae016d237",
"transactions" : [
],
"coinbaseaux" : {
"flags" : "062f503253482f"
},
"coinbasevalue" : 1588000000,
"target" : "000000000000000000000000000000000000000000000000006e439100000000",
"mintime" : 1373627219,
"mutable" : [
"time",
"transactions",
"prevblock"
],
"noncerange" : "00000000ffffffff",
"sigoplimit" : 20000,
"sizelimit" : 1000000,
"curtime" : 1373644465,
"bits" : "07ee4391",
"height" : 20203
}

So it appears that either of the interfaces is missing the prime information.

[achillez]

well whoever figures this out gets a 2.5BTC bounty + I'm going to throw in some BTC as well if the improvement is indeed 10x

[kendog77]

We maybe going at this from the wrong angle.

Luke Jr claims that BfgMiner has a 10-20x mining performance improvement over primecoind. If that is really true, it seems like most of the performance improvement should be in the prime implementation since both programs use CPU mining.

It should be feasible to take the prime implementation from BfgMiner and incorporate that into primecoind to get a similar performance improvement out of primecoind and avoid the getwork/gbt issue altogether...

[oroqen]

We maybe going at this from the wrong angle.

Luke Jr claims that BfgMiner has a 10-20x mining performance improvement over primecoind. If that is really true, it seems like most of the performance improvement should be in the prime implementation since both programs use CPU mining.

It should be feasible to take the prime implementation from BfgMiner and incorporate that into primecoind to get a similar performance improvement out of primecoind and avoid the getwork/gbt issue altogether...

Trouble with that is pools are going tobe needed sooner or later

[achillez]

any news on the bfgminer?

[Luke-Jr]

any news on the bfgminer?

I ported Chemisist's optimizations, and fixed a nasty bug making it do the same work over and over.
Working very well on unmodified primecoind testnet-in-a-box now.

[K1773R]

any news on the bfgminer?

I ported Chemisist's optimizations, and fixed a nasty bug making it do the same work over and over.
Working very well on unmodified primecoind testnet-in-a-box now.

check here: https://bitcointalk.org/index.php?topic=255358.0
also i found several blocks on testnet so it works

[achillez]

very awesome - I have this running on my systems now - will let folks know asap if this works